Scholarlink

<p>Extreme rainfall events that break previous records are occurring more frequently worldwide, leading to severe flooding and infrastructure damage. Conventional flood design approaches, based on extreme value analysis (EVA) of limited historical data, often fail to anticipate such unprecedented extremes. Here, we present a stochastic approach that leverages the Advanced Weather Generator (AWE-GEN) to simulate a large ensemble of 100-year hourly rainfall time series, explicitly accounting for internal climate variability. By excluding the record-breaking event year during calibration, we assess the ability of our proposed method to reproduce unseen record-breaking events. We evaluated this approach using data from 2703 rain stations across nine countries. Our results show that the stochastic approach captures record-breaking events more reliably than EVA, achieving success rates exceeding 85% for 3–12-hour durations at a 100-year return period threshold. This framework provides a more robust way for estimating rainfall extremes and supports the design of resilient infrastructure under deep uncertainty.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Record-breaking rainfall: a stochastic approach for its prediction

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年10月29日

查看全文
  • Mengzhu Chen,
  • Shanti Shwarup Mahto,
  • Xiaogang He,
  • Changhyun Jun,
  • Athanasios Paschalis,
  • Nadav Peleg,
  • Giuseppe Mascaro,
  • Simone Fatichi

摘要

Extreme rainfall events that break previous records are occurring more frequently worldwide, leading to severe flooding and infrastructure damage. Conventional flood design approaches, based on extreme value analysis (EVA) of limited historical data, often fail to anticipate such unprecedented extremes. Here, we present a stochastic approach that leverages the Advanced Weather Generator (AWE-GEN) to simulate a large ensemble of 100-year hourly rainfall time series, explicitly accounting for internal climate variability. By excluding the record-breaking event year during calibration, we assess the ability of our proposed method to reproduce unseen record-breaking events. We evaluated this approach using data from 2703 rain stations across nine countries. Our results show that the stochastic approach captures record-breaking events more reliably than EVA, achieving success rates exceeding 85% for 3–12-hour durations at a 100-year return period threshold. This framework provides a more robust way for estimating rainfall extremes and supports the design of resilient infrastructure under deep uncertainty.